Write a comment

Summary

 Ellis et al. address a critical need in public health surveillance by adapting molecular diagnostics for environmental pathogen monitoring. Traditional respiratory virus surveillance relies on individual clinical testing, which is expensive and logistically challenging in congregate settings like schools. Air sampling offers a promising alternative by detecting viral RNA shed by infected individuals within shared indoor spaces, potentially providing early warning of outbreaks without requiring individual participation. The study modified the CRISPR-Cas13 CARMEN Respiratory Viral Panel, originally designed for nasal swabs, to detect multiple respiratory pathogens in air samples collected from 15 schools in Wisconsin during 2023-2024. Key modifications included switching SARS-CoV-2 targets from Orf1a/b to the more abundant nucleocapsid gene, implementing the removal of PCR inhibitors to address environmental contaminants, and redesigning influenza A targets to enhance the detection of circulating strains. The modified assay (RVP_air) detected SARS-CoV-2 in 58.8% of samples that tested positive by qRT-PCR, while influenza A detection remained problematic despite further optimization efforts (RVP_air_flu). The assay successfully identified additional respiratory pathogens, including seasonal coronaviruses and respiratory syncytial virus, revealing circulation patterns that correlated temporally with clinical cases from the same schools. However, concordance with qRT-PCR remained moderate at best, with kappa scores ranging from 0.104-0.372 for SARS-CoV-2.These findings demonstrate that while CRISPR-based air surveillance can provide useful epidemiological insights about pathogen circulation patterns, significant sensitivity limitations compared to established methods constrain its practical utility for reliable outbreak detection and public health decision-making. 

Major Comments:

1. Authors state there is low biomass in air samples and “low concentrations of nucleic acids present”, but do not provide quantitative data to compare the range of viral RNA concentrations typically detected in these types of samples. 

   1. Authors should include in the methods section the total number of copies of purified RNA that were determined, and conduct serial dilutions of positive controls to determine the dynamic range of actual air sample concentrations. The authors should quantify using digital PCR, then adjust CPC levels accordingly. The 

2. The systematic determination of the limit of detection (LOD) and limit of quantification (LOQ) was not performed in this study. This is needed to establish assay performance for surveillance applications. While authors increase PCR cycles from 40 to 50 to improve the detection of targets, no quantitative studies are performed, which are necessary to establish the sensitive threshold of the assay and thus compare its performance. 

   1. Authors should conduct LOD studies using 10-fold serial dilutions of viral RNA spiked into blank air sample matrices processed through the complete proposed workflow, and include these in the Methods section and Results.

Minor Comments:

1.  Authors should exercise caution when using Kappa statistics, particularly in cases where extremely low prevalence may compromise statistical validity.  With only 26 of 347 samples (7.5%) testing qRT-PCR positive for influenza A, the kappa statistic becomes unreliable and produces misleadingly low scores, even when the negative agreement is high.  The authors report kappa values ranging from -0.029 to 0.085 for influenza A, without acknowledging that kappa performs poorly with imbalanced datasets, which limits the ability to assess true assay concordance. 

   1. Authors should instead consider reporting prevalence-adjusted kappa (PABAK) or AC1 statistics that are less sensitive to prevalence effects, or calculate positive and negative predictive values, sensitivity, and specificity as alternative performance metrics. Include prevalence bias index calculations to quantify the magnitude of bias affecting kappa interpretation. For low-prevalence targets, such as influenza A, emphasize the reporting of positive percent agreement and negative percent agreement separately, rather than the overall kappa. 

Competing interests

The author declares that they have no competing interests.

Use of Artificial Intelligence (AI)

The author declares that they used generative AI to come up with new ideas for their review.